The lifestyles of people have changed as a result of development of network and computer technologies. With the use of Internet, people no longer have to go out shopping but can buy various things online. Advertisers thus begin digital advertising during web browsing or multimedia playing processes in addition to the traditional print media (e.g. magazines, newspapers etc.) channels. As for multimedia playing process, the beginning or end of the video program is usually inserted with advertisements that are irrelevant to the video content. Irrelevancy of the advertisements to the content is not effective and may even annoy the audiences. As for web browsing processes, usually tags are used to compare ad contents and web page contents, such that when there is a textual relevancy between the two, the relevant advertisement is displayed. However, textual matching requires manual annotation, which is a time consuming process. Also, the texts produced are scarce and ambiguous, which is unfavorable for further analysis. Furthermore, the whole video program instead of individual image frames is subject to manual annotation. Thus, information conveyed by the images cannot accurately be grasped.
Therefore, there is a need for a solution that addresses the non-effectiveness of current advertisements, increases advertisement benefits of the advertisers, reduces disturbance of the advertisements to the users and other problems mentioned before.